Mostly for Scott: bean load order definition

[Rod Widdowson]

I see that in R7721 we introduce a bean (shibboleth.CustomViewContext) at
the very top of global-system.xml so that a bean of that name always exists.
Then we rely on someone redeclaring it global.xml to be what they really
want.

This therefore makes an assumption going forward that bean declaration order
matters and that its semantics will remain the same.

I wasn't aware that we had decided this but I am fine with it as a decision,
not last  since I would guess that the world would grind to a halt if Spring
changes something as that, even if it isn't documented (which I don't know).
It means in particular that I can simplify the custom bean injection in the
scripted resolvers.

I just wanted to check that this was done by conscious decision.

Rod